< draft-briscoe-aqm-dualq-coupled-00a.txt   draft-briscoe-aqm-dualq-coupled-00.txt >
Active Queue Management (aqm) K. De Schepper Active Queue Management (aqm) K. De Schepper
Internet-Draft Bell Labs Internet-Draft Bell Labs
Intended status: Standards Track B. Briscoe, Ed. Intended status: Standards Track B. Briscoe, Ed.
Expires: January 11, 2016 Independent Expires: February 8, 2016 Independent
O. Bondarenko O. Bondarenko
Simula Research Lab Simula Research Lab
I. Tsang I. Tsang
Bell Labs Bell Labs
July 10, 2015 August 07, 2015
DualQ Coupled AQM for Low Latency, Low Loss and Scalable Throughput DualQ Coupled AQM for Low Latency, Low Loss and Scalable Throughput
draft-briscoe-aqm-dualq-coupled-00 draft-briscoe-aqm-dualq-coupled-00
Abstract Abstract
Data Centre TCP (DCTCP) was designed to provide predictably low Data Centre TCP (DCTCP) was designed to provide predictably low
queuing latency, near-zero loss, and throughput scalability using queuing latency, near-zero loss, and throughput scalability using
explicit congestion notification (ECN) and an extremely simple explicit congestion notification (ECN) and an extremely simple
marking behaviour on switches. However, DCTCP does not co-exist with marking behaviour on switches. However, DCTCP does not co-exist with
existing TCP traffic---throughput starves. So, until now, DCTCP existing TCP traffic---throughput starves. So, until now, DCTCP
could only be deployed where a clean-slate environment could be could only be deployed where a clean-slate environment could be
arranged, such as in private data centres. This specification arranged, such as in private data centres. This specification
defines `DualQ Coupled Active Queue Management (AQM)' to allow defines `DualQ Coupled Active Queue Management (AQM)' to allow
scalable congestion controls like DCTCP to safely co-exist with scalable congestion controls like DCTCP to safely co-exist with
classic Internet traffic. The Coupled AQM ensures that a flow runs classic Internet traffic. The Coupled AQM ensures that a flow runs
at about the same rate whether it uses DCTCP or TCP Reno/Cubic, but at about the same rate whether it uses DCTCP or TCP Reno/Cubic, but
without inspecting transport layer flow identifiers. When tested in without inspecting transport layer flow identifiers. When tested in
a residential broadband setting, DCTCP achieved sub-millisecond a residential broadband setting, DCTCP achieved sub-millisecond
average queuing delay and zero congestion loss under a wide range of average queuing delay and zero congestion loss under a wide range of
mixes of DCTCP and `classic' broadband Internet traffic, without mixes of DCTCP and `Classic' broadband Internet traffic, without
compromising the performance of the classic traffic. The solution compromising the performance of the Classic traffic. The solution
also reduces network complexity and eliminates network configuration. also reduces network complexity and eliminates network configuration.
Status of This Memo Status of This Memo
This Internet-Draft is submitted in full conformance with the This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79. provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet- working documents as Internet-Drafts. The list of current Internet-
Drafts is at http://datatracker.ietf.org/drafts/current/. Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress." material or to cite them other than as "work in progress."
This Internet-Draft will expire on January 11, 2016. This Internet-Draft will expire on February 8, 2016.
Copyright Notice Copyright Notice
Copyright (c) 2015 IETF Trust and the persons identified as the Copyright (c) 2015 IETF Trust and the persons identified as the
document authors. All rights reserved. document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents Provisions Relating to IETF Documents
(http://trustee.ietf.org/license-info) in effect on the date of (http://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents publication of this document. Please review these documents
skipping to change at page 2, line 31 skipping to change at page 2, line 31
Table of Contents Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1. Problem and Scope . . . . . . . . . . . . . . . . . . . . 2 1.1. Problem and Scope . . . . . . . . . . . . . . . . . . . . 2
1.2. Terminology . . . . . . . . . . . . . . . . . . . . . . . 4 1.2. Terminology . . . . . . . . . . . . . . . . . . . . . . . 4
1.3. Features . . . . . . . . . . . . . . . . . . . . . . . . 5 1.3. Features . . . . . . . . . . . . . . . . . . . . . . . . 5
2. DualQ Coupled AQM Algorithm . . . . . . . . . . . . . . . . . 6 2. DualQ Coupled AQM Algorithm . . . . . . . . . . . . . . . . . 6
2.1. Coupled AQM . . . . . . . . . . . . . . . . . . . . . . . 6 2.1. Coupled AQM . . . . . . . . . . . . . . . . . . . . . . . 6
2.2. Dual Queue . . . . . . . . . . . . . . . . . . . . . . . 7 2.2. Dual Queue . . . . . . . . . . . . . . . . . . . . . . . 7
2.3. Traffic Classification . . . . . . . . . . . . . . . . . 7 2.3. Traffic Classification . . . . . . . . . . . . . . . . . 7
2.4. Normative Requirements . . . . . . . . . . . . . . . . . 8 2.4. Normative Requirements . . . . . . . . . . . . . . . . . 9
3. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 9 3. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 10
4. Security Considerations . . . . . . . . . . . . . . . . . . . 9 4. Security Considerations . . . . . . . . . . . . . . . . . . . 10
5. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 9 4.1. Overload Handling . . . . . . . . . . . . . . . . . . . . 10
6. References . . . . . . . . . . . . . . . . . . . . . . . . . 10 5. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 11
6.1. Normative References . . . . . . . . . . . . . . . . . . 10 6. References . . . . . . . . . . . . . . . . . . . . . . . . . 11
6.2. Informative References . . . . . . . . . . . . . . . . . 10 6.1. Normative References . . . . . . . . . . . . . . . . . . 11
Appendix A. Example DualQ Coupled Algorithm . . . . . . . . . . 12 6.2. Informative References . . . . . . . . . . . . . . . . . 11
Appendix B. Guidance on Controlling Throughput Equivalence . . . 17 Appendix A. Example DualQ Coupled Algorithm . . . . . . . . . . 14
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 18 Appendix B. Guidance on Controlling Throughput Equivalence . . . 20
Appendix C. DCTCP Safety Enhancements . . . . . . . . . . . . . 21
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 22
1. Introduction 1. Introduction
1.1. Problem and Scope 1.1. Problem and Scope
Latency is becoming the critical performance factor for many (most?) Latency is becoming the critical performance factor for many (most?)
applications on the public Internet, e.g. Web, voice, conversational applications on the public Internet, e.g. Web, voice, conversational
video, gaming and finance apps. In the developed world, further video, gaming and finance apps. In the developed world, further
increases in access network bit-rate offer diminishing returns, increases in access network bit-rate offer diminishing returns,
whereas latency is still a multi-facetted problem. In the last whereas latency is still a multi-faceted problem. In the last decade
decade or so, much has been done to reduce propagation time by or so, much has been done to reduce propagation time by placing
placing caches or servers closer to users. However, queuing remains caches or servers closer to users. However, queuing remains a major
a major component of latency. component of latency.
The Diffserv architecture provides Expedited Forwarding [RFC3246], so The Diffserv architecture provides Expedited Forwarding [RFC3246], so
that low latency traffic can jump the queue of other traffic. that low latency traffic can jump the queue of other traffic.
However, on access links dedicated to individual sites (homes, small However, on access links dedicated to individual sites (homes, small
enterprises or mobile devices), often all traffic at any one time enterprises or mobile devices), often all traffic at any one time
will be latency-sensitive. Then Diffserv is of little use. Instead, will be latency-sensitive. Then Diffserv is of little use. Instead,
we need to remove the causes of any unnecessary delay. we need to remove the causes of any unnecessary delay.
The bufferbloat project has shown that excessively-large buffering The bufferbloat project has shown that excessively-large buffering
(`bufferbloat') has been introducing significantly more delay than (`bufferbloat') has been introducing significantly more delay than
skipping to change at page 3, line 49 skipping to change at page 3, line 51
1. the `sawtooth' varying rate of TCP itself; 1. the `sawtooth' varying rate of TCP itself;
2. the smoothing delay deliberately introduced into AQMs to permit 2. the smoothing delay deliberately introduced into AQMs to permit
bursts without triggering losses. bursts without triggering losses.
The former causes a flow's round trip time (RTT) to vary from about 1 The former causes a flow's round trip time (RTT) to vary from about 1
to 2 times the base RTT between the machines in question. The latter to 2 times the base RTT between the machines in question. The latter
delays the system's response to change by a worst-case delays the system's response to change by a worst-case
(transcontinental) RTT, which could be hundreds of times the actual (transcontinental) RTT, which could be hundreds of times the actual
RTT of typical traffic from localised CDNs. RTT of typical traffic from localized CDNs.
Latency is not our only concern: Latency is not our only concern:
3. It was known when TCP was first developed that it would not scale 3. It was known when TCP was first developed that it would not scale
to high bandwidth-delay products. to high bandwidth-delay products.
Given regular broadband bit-rates over WAN distances are Given regular broadband bit-rates over WAN distances are
already [RFC3649] beyond the scaling range of `classic' TCP Reno, already [RFC3649] beyond the scaling range of `classic' TCP Reno,
`less unscalable' Cubic [I-D.zimmermann-tcpm-cubic] and `less unscalable' Cubic [I-D.zimmermann-tcpm-cubic] and
Compound [I-D.sridharan-tcpm-ctcp] variants of TCP have been Compound [I-D.sridharan-tcpm-ctcp] variants of TCP have been
successfully deployed. However, these are now approaching their successfully deployed. However, these are now approaching their
scaling limits. Unfortunately, fully scalable TCPs such as DCTCP scaling limits. Unfortunately, fully scalable TCPs such as DCTCP
cause `classic' TCP to starve itself, which is why they have been cause `classic' TCP to starve itself, which is why they have been
confined to private data centres or research testbeds (until now). confined to private data centres or research testbeds (until now).
This document specifies a `DualQ Coupled AQM' that solves the problem This document specifies a `DualQ Coupled AQM' that solves the problem
of coexistence between DCTCP and classic flows, without having to of coexistence between DCTCP and classic flows, without having to
inspect flow identifiers. It needs fewer operations per packet than inspect flow identifiers. The AQM is not like flow-queuing
RED uses. Also, no network configuration is needed for a wide range approaches [I-D.ietf-aqm-fq-codel] that classify packets by flow
of scenarios. Therefore it is believed it would be applicable and identifier into numerous separate queues in order to isolate sparse
easy to deploy in all types of buffers; buffers in cost-reduced mass- flows from the higher latency in the queues assigned to heavier flow.
market residential equipment; buffers in end-system stacks; buffers In contrast, the AQM exploits the behaviour of scalable congestion
in carrier-scale equipment including remote access servers, routers, controls like DCTCP so that every packet in every flow sharing the
firewalls and ethernet switches; buffers in network interface cards, queue for DCTCP-like traffic can be served with very low latency.
buffers in virtualised network appliances, hypervisors, and so on.
The AQM needs fewer operations per packet than RED uses. Also, no
network configuration is needed for a wide range of scenarios where
the range of RTTs is typical for the public Internet. Therefore it
is believed the Coupled AQM would be applicable and easy to deploy in
all types of buffers; buffers in cost-reduced mass-market residential
equipment; buffers in end-system stacks; buffers in carrier-scale
equipment including remote access servers, routers, firewalls and
Ethernet switches; buffers in network interface cards, buffers in
virtualized network appliances, hypervisors, and so on.
The supporting paper [DCttH15] gives the full rationale for the AQM's The supporting paper [DCttH15] gives the full rationale for the AQM's
design, both discursively and in more precise mathematical form. design, both discursively and in more precise mathematical form.
1.2. Terminology 1.2. Terminology
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in [RFC2119]. In this document are to be interpreted as described in [RFC2119]. In this
document, these words will appear with that interpretation only when document, these words will appear with that interpretation only when
skipping to change at page 5, line 20 skipping to change at page 5, line 34
The AQM couples marking and/or dropping across the two queues such The AQM couples marking and/or dropping across the two queues such
that a flow will get roughly the same throughput whichever it uses. that a flow will get roughly the same throughput whichever it uses.
Therefore both queues can feed into the full capacity of a link and Therefore both queues can feed into the full capacity of a link and
no rates need to be configured for the queues. The L4S queue enables no rates need to be configured for the queues. The L4S queue enables
scalable congestion controls like DCTCP to give stunningly low and scalable congestion controls like DCTCP to give stunningly low and
predictably low latency, without compromising the performance of predictably low latency, without compromising the performance of
competing 'Classic' Internet traffic. Thousands of tests have been competing 'Classic' Internet traffic. Thousands of tests have been
conducted in a typical fixed residential broadband setting. Typical conducted in a typical fixed residential broadband setting. Typical
experiments used a base round trip delay of 7ms between the data experiments used a base round trip delay of 7ms between the data
centre and home network, and large amounts of background traffic in centre and home network, and large amounts of background traffic in
both queues. For all L4S traffic, the AQM kept the 99th percentile both queues. For every L4S packet, the AQM kept the 99th percentile
of queuing delay to about 1ms, and no losses at all were introduced of queuing delay to about 1ms, and no losses at all were introduced
by the AQM. Details of the extensive experiments will be made by the AQM. Details of the extensive experiments will be made
available [DCttH15]. available [DCttH15].
Subjective testing was also conducted using a demanding panoramic Subjective testing was also conducted using a demanding panoramic
interactive video application run over a stack with DCTCP enabled and interactive video application run over a stack with DCTCP enabled and
deployed on the testbed. Each user could pan or zoom their own high deployed on the testbed. Each user could pan or zoom their own high
definition (HD) sub-window of a larger video scene from a football definition (HD) sub-window of a larger video scene from a football
match. Even though the user was also downloading large amounts of match. Even though the user was also downloading large amounts of
L4S and Classic data, latency was so low that the picture appeared to L4S and Classic data, latency was so low that the picture appeared to
stick to their finger on the touchpad (all the L4S data achieved the stick to their finger on the touchpad (all the L4S data achieved the
same ultra-low latency). With an alternative AQM, the video same ultra-low latency). With an alternative AQM, the video
noticeably lagged behind the finger gestures. noticeably lagged behind the finger gestures.
The experiments used the implementation of DCTCP that is deployed in Unlike Diffserv Expedited Forwarding, the L4S queue does not have to
private data centres, without any modification despite its known be limited to a small proportion of the link capacity in order to
deficiencies. Nonetheless, certain modifications will be necessary achieve low delay. The L4S queue can be filled with a heavy load of
before DCTCP is safe to use on the Internet, In particular it needs capacity-seeking flows like DCTCP and still achieve low delay. The
to: L4S queue does not rely on the presence of other traffic in the
Classic queue that can be 'overtaken'. It gives low latency to L4S
o negotiate its altered feedback semantics, which conveys the extent traffic whether or not there is Classic traffic, and the latency of
of ECN marking, not just its existence, and this feedback needs to Classic traffic does not suffer when a proportion of the traffic is
be robust to loss [I-D.ietf-tcpm-accecn-reqs]; L4S. The two queues are only necessary because DCTCP-like flows
cannot keep latency predictably low and keep utilization high if they
o fall back to Reno or Cubic behaviour on loss; are mixed with legacy TCP flows,
o average ECN feedback over its own RTT, not the hard-coded RTT
suitable only for data-centres, perhaps like Relentless
TCP [Mathis09];
o handle a window of less than 2 when the RTT is low, rather than
increase the queue [TCP-sub-mss-w].
Other, non-essential enhancements to DCTCP can be envisaged. The experiments used the Linux implementation of DCTCP that is
However, the focus of this specification is to get the network deployed in private data centres, without any modification despite
service in place. Then, without any management intervention, its known deficiencies. Nonetheless, certain modifications will be
applications can exploit it by migrating to scalable controls like necessary before DCTCP is safe to use on the Internet, which are
DCTCP, which can then evolve _while_ their benefits are being enjoyed recorded for now in Appendix C. However, the focus of this
by everyone on the Internet. specification is to get the network service in place. Then, without
any management intervention, applications can exploit it by migrating
to scalable controls like DCTCP, which can then evolve _while_ their
benefits are being enjoyed by everyone on the Internet.
2. DualQ Coupled AQM Algorithm 2. DualQ Coupled AQM Algorithm
There are two main aspects to the algorithm: There are two main aspects to the algorithm:
o the Coupled AQM that addresses throughput equivalence between o the Coupled AQM that addresses throughput equivalence between
Classic (e.g. Reno, Cubic) flows and L4S (e.g. DCTCP) flows Classic (e.g. Reno, Cubic) flows and L4S (e.g. DCTCP) flows
o the Dual Queue structure that provides latency separation for L4S o the Dual Queue structure that provides latency separation for L4S
flows to isolate them from the typically large Classic queue. flows to isolate them from the typically large Classic queue.
2.1. Coupled AQM 2.1. Coupled AQM
In the 1990s, the `TCP formula' was derived for the relationship In the 1990s, the `TCP formula' was derived for the relationship
between TCP's congestion window, cwnd, and its drop probability, p. between TCP's congestion window, cwnd, and its drop probability, p.
To a first order approximation, cwnd of TCP Reno is inversely To a first order approximation, cwnd of TCP Reno is inversely
proportional to the square root of p. In our supporting paper proportional to the square root of p. TCP Cubic implements a Reno-
[DCttH15], we derive the equivalent relationship for DCTCP, for which compatibility mode, which is the only relevant mode for typical RTTs
cwnd is inversely proportional to p, where in this case p is the ECN under 20ms, while the throughput of a single flow is less than about
marking probability. 500Mb/s. Therefore we can assume that Cubic traffic behaves similar
to Reno (but with a slightly different constant of proportionality),
and we shall use the term 'Classic' for the collection of Reno and
Cubic in Reno mode.
TCP Cubic implements a Reno-compatibility mode, which is the only In our supporting paper [DCttH15], we derive the equivalent rate
relevant mode for typical RTTs under 20ms, while the throughput of a equation for DCTCP, for which cwnd is inversely proportional to p
single flow is less than about 500Mb/s. Therefore we can assume that (not the square root), where in this case p is the ECN marking
Cubic traffic behaves similar to Reno (but with a slightly different probability. DCTCP is not the only congestion control that behaves
constant of proportionality). like this, so we use the term 'L4S' traffic for all similar
behaviour.
Therefore, in order to make a DCTCP flow run at roughly the same rate In order to make a DCTCP flow run at roughly the same rate as a Reno
as a Reno or Cubic TCP flow (all other factors being equal), the drop TCP flow (all other factors being equal), we make the drop
probability of Reno and Cubic has to be proportional to the square of probability for Classic traffic, p_C distinct from the marking
the ECN marking probability applied to DCTCP. It turns out there is probability for L4S traffic, p_L (in contrast to RFC3168 which
a really simple way to implement the square of a probability - by requires them to be the same). We make the Classic drop probability
testing the queue against two random numbers not one. This is the p_C proportional to the square of the L4S marking probability p_L.
approach adopted in Appendix A. This is because we need to make the Reno flow rate equal the DCTCP
flow rate, so we have to square the square root of p_C in the Reno
rate equation to make it the same as the straight p_L in the DCTCP
rate equation.
Stating this as a formula, the relation between L4S marking There is a really simple way to implement the square of a probability
probability, p_L and Classic drop probability, p_C needs to take the - by testing the queue against two random numbers not one. This is
the approach adopted in Appendix A.
Stating this as a formula, the relation between Classic drop
probability, p_C, and L4S marking probability, p_L needs to take the
form: form:
p_C = ( p_L / 2^k )^2 (1) p_C = ( p_L / 2^k )^2 (1)
where 2^k is the constant of proportionality, which is expressed as a where 2^k is the constant of proportionality, which is expressed as a
power of 2 so that implementations can avoid costly division by power of 2 so that implementations can avoid costly division by
shifting p_L by k bits to the right. shifting p_L by k bits to the right.
2.2. Dual Queue 2.2. Dual Queue
skipping to change at page 7, line 22 skipping to change at page 7, line 43
for L4S traffic, and it is scheduled with strict priority. for L4S traffic, and it is scheduled with strict priority.
Nonetheless, coupled marking ensures that giving priority to L4S Nonetheless, coupled marking ensures that giving priority to L4S
traffic still leaves the right amount of spare scheduling time for traffic still leaves the right amount of spare scheduling time for
Classic flows to each get equivalent throughput to DCTCP flows (all Classic flows to each get equivalent throughput to DCTCP flows (all
other factors such as RTT being equal). The algorithm achieves this other factors such as RTT being equal). The algorithm achieves this
without having to inspect flow identifiers. without having to inspect flow identifiers.
2.3. Traffic Classification 2.3. Traffic Classification
Both the Coupled AQM and DualQ mechanisms need an identifier to Both the Coupled AQM and DualQ mechanisms need an identifier to
distinguish L4S and C packets, which will need to be standaridsed. distinguish L4S and C packets, which will need to be standardized.
In our tests we used a cleared ECN field to indicate C packets and In our tests we used a cleared ECN field to indicate C packets and
L4S otherwise. The ECN specification [RFC3168] currently defines a L4S otherwise. The ECN specification [RFC3168] currently defines a
mark as equivalent to a drop. However, it says mark as equivalent to a drop. However, it says
"An environment where all end nodes were ECN-Capable could allow "An environment where all end nodes were ECN-Capable could allow
new criteria to be developed for setting the CE codepoint, and new new criteria to be developed for setting the CE codepoint, and new
congestion control mechanisms for end-node reaction to CE packets. congestion control mechanisms for end-node reaction to CE packets.
However, this is a research issue, and as such is not addressed in However, this is a research issue, and as such is not addressed in
this document." this document."
and [RFC4774]} gives valid ways to alter ECN's semantics without and [RFC4774]} gives valid ways to alter ECN's semantics without
harming interoperability. harming interoperability.
Since publication in 2001,deployment of RFC3168 ECN has been dogged Since publication in 2001,deployment of RFC3168 ECN has been dogged
by bugs and misunderstandings. In recent years RFC3168 ECN has been by bugs and misunderstandings. In recent years RFC3168 ECN has been
deployed quite successfully on servers [ECN_Deploy], and until deployed quite successfully on servers [ECN_Deploy], and until
recently it was deployed but not enabled on a fair proportion of user recently it was deployed but not enabled on a fair proportion of user
machines. Recently one major developer developer of client devices machines. Recently one major developer of client devices has
has finally widely deployed ECN by remote configuration, with it configured ECN on-by-default in its beta releases. However although
configured on-by-default. However although some network equipment some network equipment vendors and developers have implemented ECN,
vendors and developers have implemented ECN, there is little evidence there is little evidence that any public network operator is
that any public network operator is considering or has deployed ECN- considering or has deployed ECN-capable AQMs on network equipment
capable AQMs on network equipment yet. yet.
A number of private data centre operators have deployed ECN, but not A number of private data centre operators have deployed ECN, but not
RFC3168 ECN. Instead, they are using DCTCP to get predictable ultra- RFC3168 ECN. Instead, they are using DCTCP to get predictable ultra-
low latency, and they are either ensuring that there is no non-DCTCP low latency, and they are either ensuring that there is no non-DCTCP
traffic [I-D.bensley-tcpm-dctcp], or they are segregating such traffic [I-D.bensley-tcpm-dctcp], or they are segregating such
traffic from DCTCP using Diffserv [DCTCP_Pitfalls]. The RFC3168 traffic from DCTCP using Diffserv [DCTCP_Pitfalls]. The RFC3168
approach merely prevents drop, whereas the DCTCP approach provides approach merely prevents drop, whereas the DCTCP approach provides
scalable throughput and ultra-low latency as well as avoiding drop. scalable throughput and ultra-low latency as well as avoiding drop.
Consequently it has been questioned whether the RFC3168 approach Consequently it has been questioned whether the RFC3168 approach
offers enough performance improvement for an operator to countenance offers enough performance improvement for an operator to countenance
skipping to change at page 8, line 50 skipping to change at page 9, line 24
necessary to agree on the meaning of an ECN marking on L4S traffic, necessary to agree on the meaning of an ECN marking on L4S traffic,
relative to a drop of Classic traffic. In order to prevent relative to a drop of Classic traffic. In order to prevent
starvation of Classic traffic by scalable L4S traffic (e.g. DCTCP) starvation of Classic traffic by scalable L4S traffic (e.g. DCTCP)
the drop probability of Classic traffic MUST be proportional to the the drop probability of Classic traffic MUST be proportional to the
square of the marking probability of L4S traffic, In other words, the square of the marking probability of L4S traffic, In other words, the
power to which p_L is raised in Eqn. (1) MUST be 2. power to which p_L is raised in Eqn. (1) MUST be 2.
The constant of proportionality, k, in Eqn (1) determines the The constant of proportionality, k, in Eqn (1) determines the
relative flow rates of Classic and L4S flows when the AQM concerned relative flow rates of Classic and L4S flows when the AQM concerned
is the bottleneck (all other factors being equal). k does not have to is the bottleneck (all other factors being equal). k does not have to
be standardised because differences do not prevent interoperability. be standardized because differences do not prevent interoperability.
However, k has to take some value, and each operator can make that However, k has to take some value, and each operator can make that
choice. choice.
A value of k=0 is RECOMMENDED as the default for public Internet A value of k=0 is RECOMMENDED as the default for public Internet
access networks, assuming the DCTCP algorithm remains similar to that access networks, assuming the DCTCP algorithm remains similar to that
in [I-D.bensley-tcpm-dctcp]. Nonetheless choice of k is a matter of in [I-D.bensley-tcpm-dctcp]. Nonetheless choice of k is a matter of
operator policy, and operators MAY choose a different value using operator policy, and operators MAY choose a different value using
Table 1 and the guidelines in Appendix B. Table 1 and the guidelines in Appendix B.
Typically, access network operators isolate customers from each other Typically, access network operators isolate customers from each other
with some form of layer-2 multiplexing (TDM in DOCSIS, CDMA in 3G) or with some form of layer-2 multiplexing (TDM in DOCSIS, CDMA in 3G) or
L3 scheduling (WRR in broadband), rather than relying on TCP to share L3 scheduling (WRR in broadband), rather than relying on TCP to share
capacity between customers [RFC0970]}. In such cases, the choice of k capacity between customers [RFC0970]. In such cases, the choice of k
will solely affect relative flow rates within the customer's access will solely affect relative flow rates within the customer's access
capacity, not between customers. Also, k would not affect rates of capacity, not between customers. Also, k would not affect rates of
small flows, nor long flows at any times when they are all Classic or small flows, nor long flows at any times when they are all Classic or
all L4S. all L4S.
An example DualQ Coupled AQM algorithm is given in Appendix A. The An example DualQ Coupled AQM algorithm is given in Appendix A.
AQM in each queue is based on an AQM called Curvy RED, which is Marking and dropping in each queue is based on an AQM called Curvy
intended to improve on RED, PIE and CoDel. We have found that Curvy RED, which is intended to improve on RED, PIE and CoDel. We have
RED offers good performance, requires less operations per packet than found that Curvy RED offers good performance, requires less
RED and is insensitive to configuration. Nonetheless, it would be operations per packet than RED and is insensitive to configuration.
possible to control each queue with an alternative AQM, as long as Nonetheless, it would be possible to control each queue with an
the above normative requirements (those expressed in capitals) are alternative AQM, as long as the above normative requirements (those
observed, which are intended to be independent of the specific AQM. expressed in capitals) are observed, which are intended to be
independent of the specific AQM.
{ToDo: Add management and monitoring requirements}
3. IANA Considerations 3. IANA Considerations
This specification contains no IANA considerations. This specification contains no IANA considerations.
4. Security Considerations 4. Security Considerations
4.1. Overload Handling
Where the interests of users or flows might conflict, it could be Where the interests of users or flows might conflict, it could be
necessary to police traffic to isolate any harm to performance. This necessary to police traffic to isolate any harm to performance. This
is a policy issue that needs to be seperable from a basic AQM, but is a policy issue that needs to be separable from a basic AQM, but
the scheme does need to handle overload. A tradeoff needs to be made the scheme does need to handle overload. A trade-off needs to be
between complexity and the risk of either class harming the other. made between complexity and the risk of either class harming the
It is an operator policy to define what must happen if the service other. It is an operator policy to define what must happen if the
time of the classic queue becomes too big. Actions can include delay service time of the classic queue becomes too great. In the
based scheduling, common drop, etc. {ToDo: Expand this discussion in following subsections three optional non-exclusive overload
a future draft, with examples.} protections are defined. Their objective is for the overload
behaviour of the DualQ AQM to be similar to a single queue AQM.
Other overload protections can be envisaged:
Minimum throughput service: By replacing the priority scheduler
with a weighted round robin scheduler, a minimum throughput
service can be guaranteed for Classic traffic. Typically the
scheduling weight of the Classic queue will be small (e.g. 5%) to
avoid interference with the coupling but big enough to avoid
complete starvation of Classic traffic.
Drop on overload: On severe overload, e.g. due to non responsive
traffic, queues will typically overflow and packet drop will be
unavoidable when the queues reach their limits. The drop-limit of
each queue should be configured by specifying the maximum
supported load and determining the expected maximum size of each
queue when that load is separately applied to each queue. The
Classic queue limit will typically be larger than the L4S queue
limit. Overflow of one traffic type will automatically result in
drop in its respective queue. Both traffic types will get a high
congestion signal, due to the coupled marking, which will result
in similar starvation of responsive traffic in both queues. Thus,
the behaviour will be like a single queue AQM. To further improve
the arrival fairness of a single queue an extra overall AQM limit
can be applied, which is a limit to the sum of both queues. To be
effective, it should be configured to be less than the sum of the
limits of both queues, but greater than the maximum individual
queue limit. It ensures that the drop probability of unresponsive
traffic will be independent of its traffic type.
Delay on overload: To control milder overload of responsive traffic,
particularly when close to the maximum congestion signal, delay
can be used as an alternative congestion control mechanism. The
Dual Queue Coupled AQM can be made to behave like a single FIFO
queue with differentiated service times by replacing the priority
scheduler with a very simple "biased longest sojourn time first
scheduler". The bias is defined as a maximum sojourn time
difference (T_m) between the Classic and L4S packets. The
scheduler adds T_m to the sojourn time of the next L4S packet,
before comparing it with the timestamp of the next Classic packet,
then it selects the packet with the greater adjusted sojourn time.
This time shifted FIFO queue behaves just like a single FIFO queue
under moderate and high overload.
5. Acknowledgements 5. Acknowledgements
Thanks to Anil Agarwal for detailed review comments and suggestions
on how to make our explanation clearer.
The authors' contributions are part-funded by the European Community The authors' contributions are part-funded by the European Community
under its Seventh Framework Programme through the Reducing Internet under its Seventh Framework Programme through the Reducing Internet
Transport Latency (RITE) project (ICT-317700). The views expressed Transport Latency (RITE) project (ICT-317700). The views expressed
here are solely those of the authors. here are solely those of the authors.
6. References 6. References
6.1. Normative References 6.1. Normative References
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119, March 1997. Requirement Levels", BCP 14, RFC 2119,
DOI 10.17487/RFC2119, March 1997,
<http://www.rfc-editor.org/info/rfc2119>.
6.2. Informative References 6.2. Informative References
[ARED01] Floyd, S., Gummadi, R., and S. Shenker, "Adaptive RED: An [ARED01] Floyd, S., Gummadi, R., and S. Shenker, "Adaptive RED: An
Algorithm for Increasing the Robustness of RED's Active Algorithm for Increasing the Robustness of RED's Active
Queue Management", ACIRI Technical Report , August 2001, Queue Management", ACIRI Technical Report , August 2001,
<http://www.icir.org/floyd/red.html>. <http://www.icir.org/floyd/red.html>.
[CoDel] Nichols, K. and V. Jacobson, "Controlling Queue Delay", [CoDel] Nichols, K. and V. Jacobson, "Controlling Queue Delay",
ACM Queue 10(5), May 2012, ACM Queue 10(5), May 2012,
<http://queue.acm.org/issuedetail.cfm?issue=2208917>. <http://queue.acm.org/issuedetail.cfm?issue=2208917>.
[CRED_Insights] [CRED_Insights]
Briscoe, B., "Insights from Curvy RED (Random Early Briscoe, B., "Insights from Curvy RED (Random Early
Detection)", BT Technical Report TR-TUB8-2015-003, May Detection)", BT Technical Report TR-TUB8-2015-003, July
2015, 2015,
<http://www.bobbriscoe.net/projects/latency/credi_tr.pdf>. <http://www.bobbriscoe.net/projects/latency/credi_tr.pdf>.
[DCTCP_Pitfalls] [DCTCP_Pitfalls]
Judd, G., "Attaining the Promise and Avoiding the Pitfalls Judd, G., "Attaining the Promise and Avoiding the Pitfalls
of TCP in the Datacenter", 12th USENIX Symposium on of TCP in the Datacenter", 12th USENIX Symposium on
Networked Systems Design and Implementation (NSDI 15) Networked Systems Design and Implementation (NSDI 15)
145--157, May 2015, 145--157, May 2015,
<http://blogs.usenix.org/conference/nsdi15/technical- <http://blogs.usenix.org/conference/nsdi15/technical-
sessions/presentation/judd>. sessions/presentation/judd>.
skipping to change at page 11, line 9 skipping to change at page 12, line 37
Trammell, B., Kuehlewind, M., Boppart, D., Learmonth, I., Trammell, B., Kuehlewind, M., Boppart, D., Learmonth, I.,
Fairhurst, G., and R. Scheffenegger, "Enabling Internet- Fairhurst, G., and R. Scheffenegger, "Enabling Internet-
Wide Deployment of Explicit Congestion Notification", Proc Wide Deployment of Explicit Congestion Notification", Proc
Passive & Active Measurement (PAM'15) Conference , 2015, Passive & Active Measurement (PAM'15) Conference , 2015,
<http://ecn.ethz.ch/ecn-pam15.pdf>. <http://ecn.ethz.ch/ecn-pam15.pdf>.
[I-D.bensley-tcpm-dctcp] [I-D.bensley-tcpm-dctcp]
Bensley, S., Eggert, L., Thaler, D., Balasubramanian, P., Bensley, S., Eggert, L., Thaler, D., Balasubramanian, P.,
and G. Judd, "Microsoft's Datacenter TCP (DCTCP): TCP and G. Judd, "Microsoft's Datacenter TCP (DCTCP): TCP
Congestion Control for Datacenters", draft-bensley-tcpm- Congestion Control for Datacenters", draft-bensley-tcpm-
dctcp-04 (work in progress), July 2015. dctcp-05 (work in progress), July 2015.
[I-D.ietf-aqm-fq-codel] [I-D.ietf-aqm-fq-codel]
Hoeiland-Joergensen, T., McKenney, P., Hoeiland-Joergensen, T., McKenney, P.,
dave.taht@gmail.com, d., Gettys, J., and E. Dumazet, dave.taht@gmail.com, d., Gettys, J., and E. Dumazet,
"FlowQueue-Codel", draft-ietf-aqm-fq-codel-01 (work in "FlowQueue-Codel", draft-ietf-aqm-fq-codel-01 (work in
progress), July 2015. progress), July 2015.
[I-D.ietf-aqm-pie] [I-D.ietf-aqm-pie]
Pan, R., Natarajan, P., Baker, F., and G. White, "PIE: A Pan, R., Natarajan, P., Baker, F., and G. White, "PIE: A
Lightweight Control Scheme To Address the Bufferbloat Lightweight Control Scheme To Address the Bufferbloat
skipping to change at page 11, line 46 skipping to change at page 13, line 28
Rhee, I., Xu, L., Ha, S., Zimmermann, A., Eggert, L., and Rhee, I., Xu, L., Ha, S., Zimmermann, A., Eggert, L., and
R. Scheffenegger, "CUBIC for Fast Long-Distance Networks", R. Scheffenegger, "CUBIC for Fast Long-Distance Networks",
draft-zimmermann-tcpm-cubic-01 (work in progress), April draft-zimmermann-tcpm-cubic-01 (work in progress), April
2015. 2015.
[Mathis09] [Mathis09]
Mathis, M., "Relentless Congestion Control", PFLDNeT'09 , Mathis, M., "Relentless Congestion Control", PFLDNeT'09 ,
May 2009, <http://www.hpcc.jp/pfldnet2009/ May 2009, <http://www.hpcc.jp/pfldnet2009/
Program_files/1569198525.pdf>. Program_files/1569198525.pdf>.
[RFC0970] Nagle, J., "On packet switches with infinite storage", RFC [RFC0970] Nagle, J., "On Packet Switches With Infinite Storage",
970, December 1985. RFC 970, DOI 10.17487/RFC0970, December 1985,
<http://www.rfc-editor.org/info/rfc970>.
[RFC2309] Braden, B., Clark, D., Crowcroft, J., Davie, B., Deering, [RFC2309] Braden, B., Clark, D., Crowcroft, J., Davie, B., Deering,
S., Estrin, D., Floyd, S., Jacobson, V., Minshall, G., S., Estrin, D., Floyd, S., Jacobson, V., Minshall, G.,
Partridge, C., Peterson, L., Ramakrishnan, K., Shenker, Partridge, C., Peterson, L., Ramakrishnan, K., Shenker,
S., Wroclawski, J., and L. Zhang, "Recommendations on S., Wroclawski, J., and L. Zhang, "Recommendations on
Queue Management and Congestion Avoidance in the Queue Management and Congestion Avoidance in the
Internet", RFC 2309, April 1998. Internet", RFC 2309, DOI 10.17487/RFC2309, April 1998,
<http://www.rfc-editor.org/info/rfc2309>.
[RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition
of Explicit Congestion Notification (ECN) to IP", RFC of Explicit Congestion Notification (ECN) to IP",
3168, September 2001. RFC 3168, DOI 10.17487/RFC3168, September 2001,
<http://www.rfc-editor.org/info/rfc3168>.
[RFC3246] Davie, B., Charny, A., Bennet, J., Benson, K., Le Boudec, [RFC3246] Davie, B., Charny, A., Bennet, J., Benson, K., Le Boudec,
J., Courtney, W., Davari, S., Firoiu, V., and D. J., Courtney, W., Davari, S., Firoiu, V., and D.
Stiliadis, "An Expedited Forwarding PHB (Per-Hop Stiliadis, "An Expedited Forwarding PHB (Per-Hop
Behavior)", RFC 3246, March 2002. Behavior)", RFC 3246, DOI 10.17487/RFC3246, March 2002,
<http://www.rfc-editor.org/info/rfc3246>.
[RFC3649] Floyd, S., "HighSpeed TCP for Large Congestion Windows", [RFC3649] Floyd, S., "HighSpeed TCP for Large Congestion Windows",
RFC 3649, December 2003. RFC 3649, DOI 10.17487/RFC3649, December 2003,
<http://www.rfc-editor.org/info/rfc3649>.
[RFC4774] Floyd, S., "Specifying Alternate Semantics for the [RFC4774] Floyd, S., "Specifying Alternate Semantics for the
Explicit Congestion Notification (ECN) Field", BCP 124, Explicit Congestion Notification (ECN) Field", BCP 124,
RFC 4774, November 2006. RFC 4774, DOI 10.17487/RFC4774, November 2006,
<http://www.rfc-editor.org/info/rfc4774>.
[TCP-sub-mss-w] [TCP-sub-mss-w]
Briscoe, B. and K. De Schepper, "Scaling TCP's Congestion Briscoe, B. and K. De Schepper, "Scaling TCP's Congestion
Window for Small Round Trip Times", BT Technical Report Window for Small Round Trip Times", BT Technical Report
TR-TUB8-2015-002, May 2015, TR-TUB8-2015-002, May 2015,
<http://www.bobbriscoe.net/projects/latency/ <http://www.bobbriscoe.net/projects/latency/
sub-mss-w.pdf>. sub-mss-w.pdf>.
Appendix A. Example DualQ Coupled Algorithm Appendix A. Example DualQ Coupled Algorithm
The pseudocode below gives the DualQ Coupled AQM algorithm we used in As a concrete example, the pseudocode below gives the DualQ Coupled
testing, as a concrete example. AQM algorithm we used in testing. Although we designed the AQM to be
efficient in integer arithmetic, to aid understanding it is first
given using real-number arithmetic. Then, one possible optimization
for integer arithmetic is given, also in pseudocode. To aid
comparison, the line numbers are kept in step between the two by
using letter suffixes where the longer code needs extra lines.
1: dualq_dequeue(lq, cq) { % Couples L4S & Classic queues, lq & cq 1: dualq_dequeue(lq, cq) { % Couples L4S & Classic queues, lq & cq
2: if ( lq.dequeue(pkt) ) { 2: if ( lq.dequeue(pkt) ) {
3: if ( (lq.len() > T) || ((cq.time() << S_L) > maxrand(U)) ) { 3a: p_L = cq.sec() / 2^S_L
3b: if ( lq.byt() > T )
3c: mark(pkt)
3d: elif ( p_L > maxrand(U) )
4: mark(pkt) 4: mark(pkt)
5: } 5: return(pkt) % return the packet and stop here
6: return(pkt) % return the packet and stop here 6: }
7: } 7: while ( cq.dequeue(pkt) ) {
8: while ( cq.dequeue(pkt) ) { 8a: alpha = 2^(-f_C)
9: Q_C += (pkt.time() - Q_C) >> f_C % Classic Q EWMA 8b: Q_C = alpha * pkt.sec() + (1-alpha)* Q_C % Classic Q EWMA
10: if ( (Q_C << S_C) > maxrand(2*U) ) { 9a: sqrt_p_C = Q_C / 2^S_C
11: drop(pkt) % Squared drop, redo loop 9b: if ( sqrt_p_C > maxrand(2*U) )
12: } else { 10: drop(pkt) % Squared drop, redo loop
13: return(pkt) % return the packet and stop here 11: else
14: } 12: return(pkt) % return the packet and stop here
13: }
14: return(NULL) % no packet to dequeue
15: } 15: }
16: }
17: maxrand(u) { % return the max of u random numbers 16: maxrand(u) { % return the max of u random numbers
18: maxr=0 17: maxr=0
19: while (u-- > 0) { 18: while (u-- > 0)
20: maxr = max(maxr, rand()) 19: maxr = max(maxr, rand()) % 0 <= rand() < 1
20: return(maxr)
21: } 21: }
22: return(maxr)
23: }
Figure 1: Example Dequeue Algorithm for Coupled DualQ AQM Figure 1: Example Dequeue Pseudocode for Coupled DualQ AQM
Packet classification code is not shown, as it is no different from Packet classification code is not shown, as it is no different from
regular packet classification. Potential classification schemes are regular packet classification. Potential classification schemes are
discussed in Section 2. Overload protection code will be included in discussed in Section 2. Overload protection code will be included in
a future draft {ToDo}. a future draft {ToDo}.
At the outer level, the structure of dualQCoupled() implements strict At the outer level, the structure of dualq_dequeue() implements
priority scheduling. The code is written assuming the AQM is applied strict priority scheduling. The code is written assuming the AQM is
on dequeue (Note 1) . Every time dualQCoupled() is called, the if- applied on dequeue (Note 1) . Every time dualq_dequeue() is called,
block in lines 2-7 determines whether there is an L4S packet to the if-block in lines 2-6 determines whether there is an L4S packet
dequeue by calling lq.dequeue(pkt), and otherwise the while-block in to dequeue by calling lq.dequeue(pkt), and otherwise the while-block
lines 8-15 determines whether there is a Classic packet to dequeue, in lines 7-13 determines whether there is a Classic packet to
by calling cq.dequeue(pkt). (Note 2) dequeue, by calling cq.dequeue(pkt). (Note 2)
In the lower priority Classic queue, a while loop is used so that, if In the lower priority Classic queue, a while loop is used so that, if
the AQM determines that a classic packet should be dropped, it the AQM determines that a classic packet should be dropped, it
continues to test for classic packets deciding whether to drop each continues to test for classic packets deciding whether to drop each
until it actually forwards one. Thus, every call to dualQCoupled() until it actually forwards one. Thus, every call to dualq_dequeue()
returns one packet if at least one is present in either queue, returns one packet if at least one is present in either queue,
otherwise it returns none. (Note 3) otherwise it returns NULL at line 14. (Note 3)
Within each queue, the decision whether to drop or mark is taken as Within each queue, the decision whether to drop or mark is taken as
follows (to simplify the explanation, for now it is assumed that follows (to simplify the explanation, it is assumed that U=1):
U=1):
L4S: If the test at line 2 determines there is an L4S packet to L4S: If the test at line 2 determines there is an L4S packet to
dequeue, the test at line 3 determines whether to mark it. It is dequeue, the tests at lines 3a and 3c determine whether to mark
actually the logical OR of two tests. The first is a simple test it. The first is a simple test of whether the L4S queue (lq.byt()
of whether the L4S queue (lq.len() [bytes]) is greater than a in bytes) is greater than a step threshold T in bytes (Note 4).
simple step threshold T [bytes]. The second test is similar to The second test is similar to the random ECN marking in RED, but
the random ECN marking in RED, but with the following differences: with the following differences: i) the marking function does not
i) the marking function does not start with a plateau of zero start with a plateau of zero marking until a minimum threshold,
marking until a minimum threshold, rather the marking probability rather the marking probability starts to increase as soon as the
starts to increase as soon as the queue is positive; ii) marking queue is positive; ii) marking depends on queuing time, not bytes,
of the L4S queue does not depend on itself, it depends on the in order to scale for any link rate without being reconfigured;
queuing time of the _other_ (Classic) queue, cq.time(); iii) iii) marking of the L4S queue does not depend on itself, it
marking depends on the instantaneous queuing time (of the other depends on the queuing time of the _other_ (Classic) queue, where
queue), not a smoothed average; iv) the queue is compared with the cq.sec() is the queuing time of the packet at the head of the
maximum of U random numbers (but if U=1, this is the same as the Classic queue (zero if empty); iv) marking depends on the
single random number used in RED). instantaneous queuing time (of the other queue), not a smoothed
average; v) the queue is compared with the maximum of U random
numbers (but if U=1, this is the same as the single random number
used in RED).
Specifically, the queueing time qc.time() is multipled by the L4S Specifically, in line 3a the marking probability p_L is set to the
slope parameter S_L that determines how steeply marking Classic queueing time qc.sec() in seconds divided by the L4S
probability rises with queuing time. The slope parameter S_L is scaling parameter 2^S_L, which represents the queuing time (in
expressed as a power of 2, so that multiplication can be seconds) at which marking probability would hit 100%. Then in line
implemented as a left bit-shift (<<). Then (if U=1) the result is 3d (if U=1) the result is compared with a uniformly distributed
compared with a uniformly distributed random number, which ensures random number between 0 and 1, which ensures that marking
that marking probability will linearly increase with queueing probability will linearly increase with queueing time. The
time. scaling parameter is expressed as a power of 2 so that division
can be implemented as a right bit-shift (>>) in line 3 of the
integer variant of the pseudocode (Figure 2).
Classic: If the test at line 8 determines that there is at least one Classic: If the test at line 7 determines that there is at least one
packet to dequeue from the Classic queue, the test at line 10 Classic packet to dequeue, the test at line 9b determines whether
determines whether to drop it. But before that, line 9 updates to drop it. But before that, line 8b updates Q_C, which is an
Q_C, which is an exponentially weighted moving average (Note 4) of exponentially weighted moving average (Note 5) of the queuing time
the queuing time in the Classic queue, pkt.time() is the in the Classic queue, where pkt.sec() is the instantaneous
instantaneous Classic queueing time and f_C is the EWMA constant queueing time of the current Classic packet and alpha is the EWMA
for the classic queue, represented as an integer power of 2, so constant for the classic queue. In line 8a, alpha is represented
that the division needed to weight the moving average can be as an integer power of 2, so that in line 8 of the integer code
implemented by a right bit-shift (>>). the division needed to weight the moving average can be
implemented by a right bit-shift (>> f_C).
Line 10 implements the drop function. Drop is proportional to the Lines 9a and 9b implement the drop function. In line 9a the
square (assuming U=1) of the averaged queueing time. First the averaged queuing time Q_C is divided by the Classic scaling
averaged queuing time Q_C is multiplied by the Classic slope parameter 2^S_C, in the same way that queuing time was scaled for
parameter S_C that determines how steeply drop probability rises L4S marking. This scaled queuing time is given the variable name
with queuing time. Again, the slope parameter S_C is expressed as sqrt_p_C because it will be squared to compute Classic drop
a power of 2, so that multiplication can be implemented as a left probability, so before it is squared it is effectively the square
bit-shift (<<). The scaled queuing time is then compared with the root of the drop probability. The squaring is done by comparing
maximum of two random numbers (assuming U=1). Comparing it with it with the maximum out of two random numbers (assuming U=1).
the maximum of two, is the same as the logical `AND' of two tests, Comparing it with the maximum out of two is the same as the
which ensures drop probability rises with the square of queuing logical `AND' of two tests, which ensures drop probability rises
time. (Note 5) with the square of queuing time (Note 6). Again, the scaling
parameter is expressed as a power of 2 so that division can be
implemented as a right bit-shift in line 9 of the integer
pseudocode.
The marking/dropping functions in each queue (lines 3 & 10) are a new The marking/dropping functions in each queue (lines 3 & 9) are two
generalisation of RED called Curvy RED, motivated as follows. When cases of a new generalization of RED called Curvy RED, motivated as
we compared the performance of our AQM with fq_CoDel and PIE, we came follows. When we compared the performance of our AQM with fq_CoDel
to the conclusion that their goal of holding queuing delay to a hard and PIE, we came to the conclusion that their goal of holding queuing
target is misguided [CRED_Insights]. As the number of flows delay to a fixed target is misguided [CRED_Insights]. As the number
increases, if the AQM does not allow TCP to increase queuing delay, of flows increases, if the AQM does not allow TCP to increase queuing
it has to introduce unusually high levels of loss. Then loss rather delay, it has to introduce abnormally high levels of loss. Then loss
than queuing becomes the dominant cause of delay for short flows, due rather than queuing becomes the dominant cause of delay for short
to timeouts and tail losses. flows, due to timeouts and tail losses.
Curvy RED constrains delay with a softened target that allows some Curvy RED constrains delay with a softened target that allows some
increase in delay as load increases. This is achieved by increasing increase in delay as load increases. This is achieved by increasing
drop probability on a convex curve relative to queue growth (a square drop probability on a convex curve relative to queue growth (the
curve in the Classic queue, if U=1). Like RED, the curve hugs the square curve in the Classic queue, if U=1). Like RED, the curve hugs
zero axis while the queue is shallow. Then, as load increases, it the zero axis while the queue is shallow. Then, as load increases,
introduces a growing barrier to higher delay. But, unlike RED, it it introduces a growing barrier to higher delay. But, unlike RED, it
requires only one parameter, the slope, not three. requires only one parameter, the scaling, not three.
There follows a summary listing of the two parameters used for each There follows a summary listing of the two parameters used for each
of the two queues: of the two queues:
Classic: Classic:
S_C : The slope that scales the queuing time of the Classic S_C : The scaling factor of the dropping function scales Classic
queue into a dropping probability in the range [0,1). To make queuing times in the range [0, 2^(S_C)] seconds into a dropping
multiplication efficient, we use an integer power of two. So probability in the range [0,1]. To make division efficient, it
we define the slope of the AQM's square curve as 2^(S_C); is constrained to be an integer power of two;
f_C : To smooth the queuing time of the Classic queue and make f_C : To smooth the queuing time of the Classic queue and make
multiplication efficient, we use an integer power of two for multiplication efficient, we use a negative integer power of
the EWMA constant, which we define as 2^(-f_C). two for the dimensionless EWMA constant, which we define as
2^(-f_C).
L4S : L4S :
S_L (or k): As for the Classic queue, we define the slope of the S_L (and k): As for the Classic queue, the scaling factor of the
L4S marking function as 2^(S_L). Note that S_L = S_C + k, L4S marking function scales Classic queueing times in the range
where k is the coupling between the queues (Section 2.1). So [0, 2^(S_L)] seconds into a probability in the range [0,1].
S_L and k count as only one parameter; Note that S_L = S_C + k, where k is the coupling between the
queues (Section 2.1). So S_L and k count as only one
parameter;
T : The queue size in packets at which step threshold marking T : The queue size in bytes at which step threshold marking
starts in the L4S queue. starts in the L4S queue.
{ToDo: These are the raw parameters used within the algorithm. A
configuration front-end could accept more meaningful parameters and
convert them into these raw parameters.}
From our experiments so far, recommended values for these parameters From our experiments so far, recommended values for these parameters
are: S_C = 1; f_C = 5; T = 5 packets. The setting of k depends on are: S_C = -1; f_C = 5; T = 5 * MTU for the range of base RTTs
policy (see Section 2.4) typical on the public Internet. [CRED_Insights] explains why these
There is also a cUrviness parameter, U, which is likely to take the parameters are applicable whatever rate link this AQM implementation
same hard-coded value for all implementations, once experiments have is deployed on and how the parameters would need to be adjusted for a
determined a good value. We have solely used U=1 in our experiments scenario with a different range of RTTs (e.g. a data centre) {ToDo
so far, but results might be even better with U=2 or higher. incorporate a summary of that report into this draft}. The setting of
k depends on policy (see Section 2.4 and Appendix B respectively for
its recommended setting and guidance on alternatives).
Note that the dropping function at line 10 calls maxrand(2*U), which There is also a cUrviness parameter, U, which is a small positive
gives twice as much curviness as the the call to maxrand(U) in the integer. It is likely to take the same hard-coded value for all
implementations, once experiments have determined a good value. We
have solely used U=1 in our experiments so far, but results might be
even better with U=2 or higher.
Note that the dropping function at line 9 calls maxrand(2*U), which
gives twice as much curviness as the call to maxrand(U) in the
marking function at line 3. This is the trick that implements the marking function at line 3. This is the trick that implements the
square rule in equation (1) (Section 2.1). So, when U=1, the L4S square rule in equation (1) (Section 2.1). This is based on the fact
marking function is linear and the Classic dropping function is that, given a number X from 1 to 6, the probability that two dice
squared. When U=2, L4S is squared and Classic is quartic. And so throws will both be less than X is the square of the probability that
on. one throw will be less than X. So, when U=1, the L4S marking
function is linear and the Classic dropping function is squared. If
U=2, L4S would be a square function and Classic would be quartic.
And so on.
The maxrand(u) function in lines 17-22 simply generates u random The maxrand(u) function in lines 16-21 simply generates u random
numbers and returns the maximum. Typically, maxrand(u) could be run numbers and returns the maximum (Note 7). Typically, maxrand(u)
in parallel out of band. For instance, if U=1, the Classic queue could be run in parallel out of band. For instance, if U=1, the
would require the maximum of two random numbers. So, the maximum of Classic queue would require the maximum of two random numbers. So,
every pair of values from a pseudorandom number generator could be instead of calling maxrand(2*U) in-band, the maximum of every pair of
generated out-of-band, and held in a buffer ready for the Classic values from a pseudorandom number generator could be generated out-
queue to consume, instead of calling maxrand(2*U) in-band. of-band, and held in a buffer ready for the Classic queue to consume.
1: dualq_dequeue(lq, cq) { % Couples L4S & Classic queues, lq & cq
2: if ( lq.dequeue(pkt) ) {
3: if ((lq.byt() > T) || ((cq.ns() >> (S_L-2)) > maxrand(U)))
4: mark(pkt)
5: return(pkt) % return the packet and stop here
6: }
7: while ( cq.dequeue(pkt) ) {
8: Q_C += (pkt.ns() - Q_C) >> f_C % Classic Q EWMA
9: if ( (Q_C >> (S_C-2) ) > maxrand(2*U) )
10: drop(pkt) % Squared drop, redo loop
11: else
12: return(pkt) % return the packet and stop here
13: }
14: return(NULL) % no packet to dequeue
15: }
Figure 2: Optimised Example Dequeue Pseudocode for Coupled DualQ AQM
using Integer Arithmetic
Notes: Notes:
1. The drain rate of the queue can vary if it is scheduled relative 1. The drain rate of the queue can vary if it is scheduled relative
to other queues, or to cater for fluctuations in a wireless to other queues, or to cater for fluctuations in a wireless
medium. To auto-adjust to changes in drain rate, the queue must medium. To auto-adjust to changes in drain rate, the queue must
be measured in time, not bytes or packets [CoDel]. In our Linux be measured in time, not bytes or packets [CoDel]. In our Linux
implementation, it was easiest to measure queuing time at implementation, it was easiest to measure queuing time at
dequeue. Queuing time can be estimated when a packet is enqueued dequeue. Queuing time can be estimated when a packet is enqueued
by measuring the queue length in bytes and dividing by the recent by measuring the queue length in bytes and dividing by the recent
drain rate. drain rate.
2. An implementation has to use priority queueing, but it need not 2. An implementation has to use priority queueing, but it need not
implement strict priority. implement strict priority.
3. If packets can be enqueued while processinging dequeue code, an 3. If packets can be enqueued while processing dequeue code, an
implementer might prefer to place the while loop around both implementer might prefer to place the while loop around both
queues so that it goes back to test again whether any L4S packets queues so that it goes back to test again whether any L4S packets
arrived while it was dropping a Classic packet. arrived while it was dropping a Classic packet.
4. An EWMA is only one possible way to filter bursts; other more 4. In order not to change too many factors at once, for now, we keep
adaptive smoothing methods could be valid. the marking function for DCTCP-only traffic as similar as
possible to DCTCP. However, unlike DCTCP, all processing is at
dequeue, so we determine whether to mark a packet at the head of
the queue by the byte-length of the queue _behind_ it. We plan
to test whether using queuing time will work in all
circumstances, and if we find that the step can cause
oscillations, we will investigate replacing it with a steep
random marking curve.
5. In practice at line 11 the Classic queue would probably test for 5. An EWMA is only one possible way to filter bursts; other more
adaptive smoothing methods could be valid and it might be
appropriate to decrease the EWMA faster than it increases.
6. In practice at line 10 the Classic queue would probably test for
ECN capability on the packet to determine whether to drop or mark ECN capability on the packet to determine whether to drop or mark
the packet. However, for brevity such detail is omitted. All the packet. However, for brevity such detail is omitted. All
packets classified into the L4S queue have to be ECN-capable, so packets classified into the L4S queue have to be ECN-capable, so
no dropping logic is necessary at line 3. Nonetheless, L4S no dropping logic is necessary at line 3. Nonetheless, L4S
packets could be dropped by overload code. packets could be dropped by overload code (see Section 4.1).
7. In the integer variant of the pseudocode (Figure 2) real numbers
are all represented as integers scaled up by 2^32. In lines 3 &
9 the function maxrand() is arranged to return an integer in the
range 0 <= maxrand() < 2^32. Queuing times are also scaled up by
2^32, but in two stages: i) In lines 3 and 8 queuing times
cq.ns() and pkt.ns() are returned in integer nanoseconds, making
the values about 2^30 times larger than when the units were
seconds, ii) then in lines 3 and 9 an adjustment of -2 to the
right bit-shift multiplies the result by 2^2, to complete the
scaling by 2^32.
Appendix B. Guidance on Controlling Throughput Equivalence Appendix B. Guidance on Controlling Throughput Equivalence
+---------------+------+-------+ +---------------+------+-------+
| RTT_C / RTT_L | Reno | Cubic | | RTT_C / RTT_L | Reno | Cubic |
+---------------+------+-------+ +---------------+------+-------+
| 1 | k=1 | k=0 | | 1 | k=1 | k=0 |
| 2 | k=2 | k=1 | | 2 | k=2 | k=1 |
| 3 | k=2 | k=2 | | 3 | k=2 | k=2 |
| 4 | k=3 | k=2 | | 4 | k=3 | k=2 |
skipping to change at page 17, line 33 skipping to change at page 21, line 7
whether it wants DCTCP flows to have roughly equal throughput with whether it wants DCTCP flows to have roughly equal throughput with
Reno or with Cubic (because, even in its Reno-compatibility mode, Reno or with Cubic (because, even in its Reno-compatibility mode,
Cubic is about 1.4 times more aggressive than Reno). Then the Cubic is about 1.4 times more aggressive than Reno). Then the
operator needs to decide at what ratio of RTTs it wants DCTCP and operator needs to decide at what ratio of RTTs it wants DCTCP and
Classic flows to have roughly equal throughput. For example choosing Classic flows to have roughly equal throughput. For example choosing
the recommended value of k=0 will make DCTCP throughput roughly the the recommended value of k=0 will make DCTCP throughput roughly the
same as Cubic, _if their RTTs are the same_. same as Cubic, _if their RTTs are the same_.
However, even if the base RTTs are the same, the actual RTTs are However, even if the base RTTs are the same, the actual RTTs are
unlikely to be the same, because Classic (Cubic or Reno) traffic unlikely to be the same, because Classic (Cubic or Reno) traffic
needs a large queue to avoid under-utilisation and excess drop, needs a large queue to avoid under-utilization and excess drop,
whereas L4S (DCTCP) does not. The operator might still choose this whereas L4S (DCTCP) does not. The operator might still choose this
policy if it judges that DCTCP throughput should be rewarded for policy if it judges that DCTCP throughput should be rewarded for
keeping its own queue short. keeping its own queue short.
On the other hand, the operator will choose one of the higher values On the other hand, the operator will choose one of the higher values
for k, if it wants to slow DCTCP down to roughly the same throughput for k, if it wants to slow DCTCP down to roughly the same throughput
as Classic flows, to compensate for Classic flows slowing themselves as Classic flows, to compensate for Classic flows slowing themselves
down by causing themselves extra queuing delay. down by causing themselves extra queuing delay.
The values for k in the table are derived from the formulae, which The values for k in the table are derived from the formulae, which
was developed in [DCttH15]: was developed in [DCttH15]:
2^k = 1.64 (RTT_reno / RTT_dc) (2) 2^k = 1.64 (RTT_reno / RTT_dc) (2)
2^k = 1.19 (RTT_cubic / RTT_dc ) (3) 2^k = 1.19 (RTT_cubic / RTT_dc ) (3)
For localised traffic from a particular ISP's data centre, we used For localized traffic from a particular ISP's data centre, we used
the measured RTTs to calculate that a value of k=3 would achieve the measured RTTs to calculate that a value of k=3 would achieve
throughput equivalence, and our experiments verified the formula very throughput equivalence, and our experiments verified the formula very
closely. closely.
Appendix C. DCTCP Safety Enhancements
This Appendix is informational not normative. It records changes
needed to DCTCP implementations so they can co-exist safely alongside
other traffic sources. They are recorded here until a more
appropriate draft is available to hold them.
Proposed changes are listed in rough order of criticality. Therefore
those later in the list may not be necessary:
o Negotiate its altered feedback semantics, which conveys the extent
of ECN marking, not just its existence, and this feedback needs to
be robust to loss [I-D.ietf-tcpm-accecn-reqs];
o fall back to Reno or Cubic behaviour on loss;
o use a packet identifier associated with the L4S service;
o average ECN feedback over its own RTT, not the hard-coded RTT
suitable only for data-centres, perhaps like Relentless
TCP [Mathis09];
o handle a window of less than 2 when the RTT is low, rather than
increase the queue [TCP-sub-mss-w].
o test heuristically whether ECN marking is emanating from an
RFC3168 AQM.
Other, non-essential enhancements to DCTCP can be envisaged.
Authors' Addresses Authors' Addresses
Koen De Schepper Koen De Schepper
Bell Labs Bell Labs
Antwerp Antwerp
Belgium Belgium
Email: koen.de_schepper@alcatel-lucent.com Email: koen.de_schepper@alcatel-lucent.com
URI: https://www.bell-labs.com/researchers/638/ URI: https://www.bell-labs.com/usr/koen.de_schepper
Bob Briscoe (editor) Bob Briscoe (editor)
Independent Independent
Email: ietf@bobbriscoe.net Email: ietf@bobbriscoe.net
URI: http://bobbriscoe.net/ URI: http://bobbriscoe.net/
Olga Bondarenko Olga Bondarenko
Simula Research Lab Simula Research Lab
Lysaker Lysaker
 End of changes. 66 change blocks. 
233 lines changed or deleted 414 lines changed or added

This html diff was produced by rfcdiff 1.42. The latest version is available from http://tools.ietf.org/tools/rfcdiff/
X-Generator: pyht 0.35